Merge branch 'main' of github.com:alibaba-damo-academy/FunASR merge

游雁

2024-05-30 7e5f987c5dffdb628a5cd904ef420e593f5ceb22

 funasr/auto/auto_model.py

@@ -1,3 +1,8 @@
#!/usr/bin/env python3
# -*- encoding: utf-8 -*-
# Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.
#  MIT License  (https://opensource.org/licenses/MIT)

import json
import time
import copy
@@ -9,66 +14,71 @@
import numpy as np
from tqdm import tqdm

from funasr.utils.misc import deep_update
from funasr.register import tables
from funasr.utils.load_utils import load_bytes
from funasr.download.file import download_from_url
from funasr.utils.timestamp_tools import timestamp_sentence
from funasr.download.download_from_hub import download_model
from funasr.utils.vad_utils import slice_padding_audio_samples
from funasr.utils.vad_utils import merge_vad
from funasr.utils.load_utils import load_audio_text_image_video
from funasr.train_utils.set_all_random_seed import set_all_random_seed
from funasr.train_utils.load_pretrained_model import load_pretrained_model
from funasr.utils.load_utils import load_audio_text_image_video
from funasr.utils.timestamp_tools import timestamp_sentence
from funasr.models.campplus.utils import sv_chunk, postprocess, distribute_spk
from funasr.utils import export_utils
from funasr.utils import misc

try:
    from funasr.models.campplus.utils import sv_chunk, postprocess, distribute_spk
    from funasr.models.campplus.cluster_backend import ClusterBackend
except:
    print("If you want to use the speaker diarization, please `pip install hdbscan`")
import pdb
    pass


def prepare_data_iterator(data_in, input_len=None, data_type=None, key=None):
    """
    
    :param input:
    :param input_len:
    :param data_type:
    :param frontend:
    :return:
    """
    """ """
    data_list = []
    key_list = []
    filelist = [".scp", ".txt", ".json", ".jsonl"]
    
    filelist = [".scp", ".txt", ".json", ".jsonl", ".text"]

    chars = string.ascii_letters + string.digits
    if isinstance(data_in, str) and data_in.startswith('http'): # url
        data_in = download_from_url(data_in)
    pdb.set_trace()
    if isinstance(data_in, str) and os.path.exists(data_in): # wav_path; filelist: wav.scp, file.jsonl;text.txt;
    if isinstance(data_in, str):
        if data_in.startswith("http://") or data_in.startswith("https://"):  # url
            data_in = download_from_url(data_in)

    if isinstance(data_in, str) and os.path.exists(
        data_in
    ):  # wav_path; filelist: wav.scp, file.jsonl;text.txt;
        _, file_extension = os.path.splitext(data_in)
        file_extension = file_extension.lower()
        if file_extension in filelist: #filelist: wav.scp, file.jsonl;text.txt;
            with open(data_in, encoding='utf-8') as fin:
        if file_extension in filelist:  # filelist: wav.scp, file.jsonl;text.txt;
            with open(data_in, encoding="utf-8") as fin:
                for line in fin:
                    key = "rand_key_" + ''.join(random.choice(chars) for _ in range(13))
                    if data_in.endswith(".jsonl"): #file.jsonl: json.dumps({"source": data})
                    key = "rand_key_" + "".join(random.choice(chars) for _ in range(13))
                    if data_in.endswith(".jsonl"):  # file.jsonl: json.dumps({"source": data})
                        lines = json.loads(line.strip())
                        data = lines["source"]
                        key = data["key"] if "key" in data else key
                    else: # filelist, wav.scp, text.txt: id \t data or data
                    else:  # filelist, wav.scp, text.txt: id \t data or data
                        lines = line.strip().split(maxsplit=1)
                        data = lines[1] if len(lines)>1 else lines[0]
                        key = lines[0] if len(lines)>1 else key
                    
                        data = lines[1] if len(lines) > 1 else lines[0]
                        key = lines[0] if len(lines) > 1 else key

                    data_list.append(data)
                    key_list.append(key)
        else:
            key = "rand_key_" + ''.join(random.choice(chars) for _ in range(13))
            if key is None:
                # key = "rand_key_" + "".join(random.choice(chars) for _ in range(13))
                key = misc.extract_filename_without_extension(data_in)
            data_list = [data_in]
            key_list = [key]
    elif isinstance(data_in, (list, tuple)):
        if data_type is not None and isinstance(data_type, (list, tuple)): # mutiple inputs
        if data_type is not None and isinstance(data_type, (list, tuple)):  # mutiple inputs
            data_list_tmp = []
            for data_in_i, data_type_i in zip(data_in, data_type):
                key_list, data_list_i = prepare_data_iterator(data_in=data_in_i, data_type=data_type_i)
                key_list, data_list_i = prepare_data_iterator(
                    data_in=data_in_i, data_type=data_type_i
                )
                data_list_tmp.append(data_list_i)
            data_list = []
            for item in zip(*data_list_tmp):
@@ -76,55 +86,72 @@
        else:
            # [audio sample point, fbank, text]
            data_list = data_in
            key_list = ["rand_key_" + ''.join(random.choice(chars) for _ in range(13)) for _ in range(len(data_in))]
    else: # raw text; audio sample point, fbank; bytes
        if isinstance(data_in, bytes): # audio bytes
            key_list = []
            for data_i in data_in:
                if isinstance(data_i, str) and os.path.exists(data_i):
                    key = misc.extract_filename_without_extension(data_i)
                else:
                    key = "rand_key_" + "".join(random.choice(chars) for _ in range(13))
                key_list.append(key)

    else:  # raw text; audio sample point, fbank; bytes
        if isinstance(data_in, bytes):  # audio bytes
            data_in = load_bytes(data_in)
        if key is None:
            key = "rand_key_" + ''.join(random.choice(chars) for _ in range(13))
            key = "rand_key_" + "".join(random.choice(chars) for _ in range(13))
        data_list = [data_in]
        key_list = [key]
    

    return key_list, data_list


class AutoModel:
    

    def __init__(self, **kwargs):
        if not kwargs.get("disable_log", False):

        log_level = getattr(logging, kwargs.get("log_level", "INFO").upper())
        logging.basicConfig(level=log_level)

        if not kwargs.get("disable_log", True):
            tables.print()
        

        model, kwargs = self.build_model(**kwargs)
        

        # if vad_model is not None, build vad model else None
        vad_model = kwargs.get("vad_model", None)
        vad_kwargs = kwargs.get("vad_model_revision", None)
        vad_kwargs = {} if kwargs.get("vad_kwargs", {}) is None else kwargs.get("vad_kwargs", {})
        if vad_model is not None:
            logging.info("Building VAD model.")
            vad_kwargs = {"model": vad_model, "model_revision": vad_kwargs, "device": kwargs["device"]}
            vad_kwargs["model"] = vad_model
            vad_kwargs["model_revision"] = kwargs.get("vad_model_revision", "master")
            vad_kwargs["device"] = kwargs["device"]
            vad_model, vad_kwargs = self.build_model(**vad_kwargs)

        # if punc_model is not None, build punc model else None
        punc_model = kwargs.get("punc_model", None)
        punc_kwargs = kwargs.get("punc_model_revision", None)
        punc_kwargs = {} if kwargs.get("punc_kwargs", {}) is None else kwargs.get("punc_kwargs", {})
        if punc_model is not None:
            logging.info("Building punc model.")
            punc_kwargs = {"model": punc_model, "model_revision": punc_kwargs, "device": kwargs["device"]}
            punc_kwargs["model"] = punc_model
            punc_kwargs["model_revision"] = kwargs.get("punc_model_revision", "master")
            punc_kwargs["device"] = kwargs["device"]
            punc_model, punc_kwargs = self.build_model(**punc_kwargs)

        # if spk_model is not None, build spk model else None
        spk_model = kwargs.get("spk_model", None)
        spk_kwargs = kwargs.get("spk_model_revision", None)
        spk_kwargs = {} if kwargs.get("spk_kwargs", {}) is None else kwargs.get("spk_kwargs", {})
        if spk_model is not None:
            logging.info("Building SPK model.")
            spk_kwargs = {"model": spk_model, "model_revision": spk_kwargs, "device": kwargs["device"]}
            spk_kwargs["model"] = spk_model
            spk_kwargs["model_revision"] = kwargs.get("spk_model_revision", "master")
            spk_kwargs["device"] = kwargs["device"]
            spk_model, spk_kwargs = self.build_model(**spk_kwargs)
            self.cb_model = ClusterBackend().to(kwargs["device"])
            spk_mode = kwargs.get("spk_mode", 'punc_segment')
            spk_mode = kwargs.get("spk_mode", "punc_segment")
            if spk_mode not in ["default", "vad_segment", "punc_segment"]:
                logging.error("spk_mode should be one of default, vad_segment and punc_segment.")
            self.spk_mode = spk_mode
            

        self.kwargs = kwargs
        self.model = model
        self.vad_model = vad_model
@@ -134,13 +161,13 @@
        self.spk_model = spk_model
        self.spk_kwargs = spk_kwargs
        self.model_path = kwargs.get("model_path")
        

    def build_model(self, **kwargs):
        assert "model" in kwargs
        if "model_conf" not in kwargs:
            logging.info("download models from model hub: {}".format(kwargs.get("model_hub", "ms")))
            logging.info("download models from model hub: {}".format(kwargs.get("hub", "ms")))
            kwargs = download_model(**kwargs)
        

        set_all_random_seed(kwargs.get("seed", 0))

        device = kwargs.get("device", "cuda")
@@ -148,79 +175,100 @@
            device = "cpu"
            kwargs["batch_size"] = 1
        kwargs["device"] = device
        
        if kwargs.get("ncpu", None):
            torch.set_num_threads(kwargs.get("ncpu"))
        

        torch.set_num_threads(kwargs.get("ncpu", 4))

        # build tokenizer
        tokenizer = kwargs.get("tokenizer", None)
        if tokenizer is not None:
            tokenizer_class = tables.tokenizer_classes.get(tokenizer)
            tokenizer = tokenizer_class(**kwargs["tokenizer_conf"])
            kwargs["tokenizer"] = tokenizer
            kwargs["token_list"] = tokenizer.token_list
            vocab_size = len(tokenizer.token_list)
            tokenizer = tokenizer_class(**kwargs.get("tokenizer_conf", {}))
            kwargs["token_list"] = (
                tokenizer.token_list if hasattr(tokenizer, "token_list") else None
            )
            kwargs["token_list"] = (
                tokenizer.get_vocab() if hasattr(tokenizer, "get_vocab") else kwargs["token_list"]
            )
            vocab_size = len(kwargs["token_list"]) if kwargs["token_list"] is not None else -1
            if vocab_size == -1 and hasattr(tokenizer, "get_vocab_size"):
                vocab_size = tokenizer.get_vocab_size()
        else:
            vocab_size = -1
        kwargs["tokenizer"] = tokenizer

        # build frontend
        frontend = kwargs.get("frontend", None)

        kwargs["input_size"] = None
        if frontend is not None:
            frontend_class = tables.frontend_classes.get(frontend)
            frontend = frontend_class(**kwargs["frontend_conf"])
            kwargs["frontend"] = frontend
            kwargs["input_size"] = frontend.output_size()

            frontend = frontend_class(**kwargs.get("frontend_conf", {}))
            kwargs["input_size"] = (
                frontend.output_size() if hasattr(frontend, "output_size") else None
            )
        kwargs["frontend"] = frontend
        # build model
        model_class = tables.model_classes.get(kwargs["model"])
        model = model_class(**kwargs, **kwargs["model_conf"], vocab_size=vocab_size)
        model_conf = {}
        deep_update(model_conf, kwargs.get("model_conf", {}))
        deep_update(model_conf, kwargs)
        model = model_class(**model_conf, vocab_size=vocab_size)
        model.to(device)
        

        # init_param
        init_param = kwargs.get("init_param", None)
        if init_param is not None:
            logging.info(f"Loading pretrained params from {init_param}")
            load_pretrained_model(
                model=model,
                path=init_param,
                ignore_init_mismatch=kwargs.get("ignore_init_mismatch", False),
                oss_bucket=kwargs.get("oss_bucket", None),
                scope_map=kwargs.get("scope_map", None),
                excludes=kwargs.get("excludes", None),
            )
        
            if os.path.exists(init_param):
                logging.info(f"Loading pretrained params from {init_param}")
                load_pretrained_model(
                    model=model,
                    path=init_param,
                    ignore_init_mismatch=kwargs.get("ignore_init_mismatch", True),
                    oss_bucket=kwargs.get("oss_bucket", None),
                    scope_map=kwargs.get("scope_map", []),
                    excludes=kwargs.get("excludes", None),
                )
            else:
                print(f"error, init_param does not exist!: {init_param}")

        # fp16
        if kwargs.get("fp16", False):
            model.to(torch.float16)
        return model, kwargs
    

    def __call__(self, *args, **cfg):
        kwargs = self.kwargs
        kwargs.update(cfg)
        deep_update(kwargs, cfg)
        res = self.model(*args, kwargs)
        return res

    def generate(self, input, input_len=None, **cfg):
        if self.vad_model is None:
            return self.inference(input, input_len=input_len, **cfg)
    

        else:
            return self.inference_with_vad(input, input_len=input_len, **cfg)
        

    def inference(self, input, input_len=None, model=None, kwargs=None, key=None, **cfg):
        kwargs = self.kwargs if kwargs is None else kwargs
        kwargs.update(cfg)
        deep_update(kwargs, cfg)
        model = self.model if model is None else model
        model.eval()

        batch_size = kwargs.get("batch_size", 1)
        # if kwargs.get("device", "cpu") == "cpu":
        #     batch_size = 1
        
        key_list, data_list = prepare_data_iterator(input, input_len=input_len, data_type=kwargs.get("data_type", None), key=key)

        key_list, data_list = prepare_data_iterator(
            input, input_len=input_len, data_type=kwargs.get("data_type", None), key=key
        )

        speed_stats = {}
        asr_result_list = []
        num_samples = len(data_list)
        disable_pbar = kwargs.get("disable_pbar", False)
        pbar = tqdm(colour="blue", total=num_samples, dynamic_ncols=True) if not disable_pbar else None
        disable_pbar = self.kwargs.get("disable_pbar", False)
        pbar = (
            tqdm(colour="blue", total=num_samples, dynamic_ncols=True) if not disable_pbar else None
        )
        time_speech_total = 0.0
        time_escape_total = 0.0
        for beg_idx in range(0, num_samples, batch_size):
@@ -229,15 +277,18 @@
            key_batch = key_list[beg_idx:end_idx]
            batch = {"data_in": data_batch, "key": key_batch}

            if (end_idx - beg_idx) == 1 and kwargs.get("data_type", None) == "fbank": # fbank
            if (end_idx - beg_idx) == 1 and kwargs.get("data_type", None) == "fbank":  # fbank
                batch["data_in"] = data_batch[0]
                batch["data_lengths"] = input_len
        

            time1 = time.perf_counter()
            with torch.no_grad():
                results, meta_data = model.inference(**batch, **kwargs)
                res = model.inference(**batch, **kwargs)
                if isinstance(res, (list, tuple)):
                    results = res[0] if len(res) > 0 else [{"text": ""}]
                    meta_data = res[1] if len(res) > 1 else {}
            time2 = time.perf_counter()
            

            asr_result_list.extend(results)

            # batch_data_time = time_per_frame_s * data_batch_i["speech_lengths"].sum().item()
@@ -248,9 +299,7 @@
            speed_stats["forward"] = f"{time_escape:0.3f}"
            speed_stats["batch_size"] = f"{len(results)}"
            speed_stats["rtf"] = f"{(time_escape) / batch_data_time:0.3f}"
            description = (
                f"{speed_stats}, "
            )
            description = f"{speed_stats}, "
            if pbar:
                pbar.update(1)
                pbar.set_description(description)
@@ -262,98 +311,128 @@
            pbar.set_description(f"rtf_avg: {time_escape_total/time_speech_total:0.3f}")
        torch.cuda.empty_cache()
        return asr_result_list
    
    def inference_with_vad(self, input, input_len=None, **cfg):
        
        # step.1: compute the vad model
        self.vad_kwargs.update(cfg)
        beg_vad = time.time()
        res = self.inference(input, input_len=input_len, model=self.vad_model, kwargs=self.vad_kwargs, **cfg)
        end_vad = time.time()
        print(f"time cost vad: {end_vad - beg_vad:0.3f}")

    def inference_with_vad(self, input, input_len=None, **cfg):
        kwargs = self.kwargs
        # step.1: compute the vad model
        deep_update(self.vad_kwargs, cfg)
        beg_vad = time.time()
        res = self.inference(
            input, input_len=input_len, model=self.vad_model, kwargs=self.vad_kwargs, **cfg
        )
        end_vad = time.time()

        #  FIX(gcf): concat the vad clips for sense vocie model for better aed
        if kwargs.get("merge_vad", False):
            for i in range(len(res)):
                res[i]["value"] = merge_vad(res[i]["value"], kwargs.get("merge_length", 15000))

        # step.2 compute asr model
        model = self.model
        kwargs = self.kwargs
        kwargs.update(cfg)
        batch_size = int(kwargs.get("batch_size_s", 300))*1000
        batch_size_threshold_ms = int(kwargs.get("batch_size_threshold_s", 60))*1000
        deep_update(kwargs, cfg)
        batch_size = max(int(kwargs.get("batch_size_s", 300)) * 1000, 1)
        batch_size_threshold_ms = int(kwargs.get("batch_size_threshold_s", 60)) * 1000
        kwargs["batch_size"] = batch_size
        
        key_list, data_list = prepare_data_iterator(input, input_len=input_len, data_type=kwargs.get("data_type", None))

        key_list, data_list = prepare_data_iterator(
            input, input_len=input_len, data_type=kwargs.get("data_type", None)
        )
        results_ret_list = []
        time_speech_total_all_samples = 1e-6

        beg_total = time.time()
        pbar_total = tqdm(colour="red", total=len(res), dynamic_ncols=True)
        pbar_total = (
            tqdm(colour="red", total=len(res), dynamic_ncols=True)
            if not kwargs.get("disable_pbar", False)
            else None
        )
        for i in range(len(res)):
            key = res[i]["key"]
            vadsegments = res[i]["value"]
            input_i = data_list[i]
            speech = load_audio_text_image_video(input_i, fs=kwargs["frontend"].fs, audio_fs=kwargs.get("fs", 16000))
            fs = kwargs["frontend"].fs if hasattr(kwargs["frontend"], "fs") else 16000
            speech = load_audio_text_image_video(input_i, fs=fs, audio_fs=kwargs.get("fs", 16000))
            speech_lengths = len(speech)
            n = len(vadsegments)
            data_with_index = [(vadsegments[i], i) for i in range(n)]
            sorted_data = sorted(data_with_index, key=lambda x: x[0][1] - x[0][0])
            results_sorted = []
            

            if not len(sorted_data):
                results_ret_list.append({"key": key, "text": "", "timestamp": []})
                logging.info("decoding, utt: {}, empty speech".format(key))
                continue

            if len(sorted_data) > 0 and len(sorted_data[0]) > 0:
                batch_size = max(batch_size, sorted_data[0][0][1] - sorted_data[0][0][0])
            
            batch_size_ms_cum = 0

            beg_idx = 0
            beg_asr_total = time.time()
            time_speech_total_per_sample = speech_lengths/16000
            time_speech_total_per_sample = speech_lengths / 16000
            time_speech_total_all_samples += time_speech_total_per_sample

            # pbar_sample = tqdm(colour="blue", total=n, dynamic_ncols=True)

            all_segments = []
            max_len_in_batch = 0
            end_idx = 1
            for j, _ in enumerate(range(0, n)):
                # pbar_sample.update(1)
                batch_size_ms_cum += (sorted_data[j][0][1] - sorted_data[j][0][0])
                if j < n - 1 and (
                    batch_size_ms_cum + sorted_data[j + 1][0][1] - sorted_data[j + 1][0][0]) < batch_size and (
                    sorted_data[j + 1][0][1] - sorted_data[j + 1][0][0]) < batch_size_threshold_ms:
                sample_length = sorted_data[j][0][1] - sorted_data[j][0][0]
                potential_batch_length = max(max_len_in_batch, sample_length) * (j + 1 - beg_idx)
                # batch_size_ms_cum += sorted_data[j][0][1] - sorted_data[j][0][0]
                if (
                    j < n - 1
                    and sample_length < batch_size_threshold_ms
                    and potential_batch_length < batch_size
                ):
                    max_len_in_batch = max(max_len_in_batch, sample_length)
                    end_idx += 1
                    continue
                batch_size_ms_cum = 0
                end_idx = j + 1
                speech_j, speech_lengths_j = slice_padding_audio_samples(speech, speech_lengths, sorted_data[beg_idx:end_idx])       
                results = self.inference(speech_j, input_len=None, model=model, kwargs=kwargs, disable_pbar=True, **cfg)

                speech_j, speech_lengths_j = slice_padding_audio_samples(
                    speech, speech_lengths, sorted_data[beg_idx:end_idx]
                )
                results = self.inference(
                    speech_j, input_len=None, model=model, kwargs=kwargs, **cfg
                )
                if self.spk_model is not None:
                    # compose vad segments: [[start_time_sec, end_time_sec, speech], [...]]
                    for _b in range(len(speech_j)):
                        vad_segments = [[sorted_data[beg_idx:end_idx][_b][0][0]/1000.0,
                                        sorted_data[beg_idx:end_idx][_b][0][1]/1000.0,
                                        np.array(speech_j[_b])]]
                        vad_segments = [
                            [
                                sorted_data[beg_idx:end_idx][_b][0][0] / 1000.0,
                                sorted_data[beg_idx:end_idx][_b][0][1] / 1000.0,
                                np.array(speech_j[_b]),
                            ]
                        ]
                        segments = sv_chunk(vad_segments)
                        all_segments.extend(segments)
                        speech_b = [i[2] for i in segments]
                        spk_res = self.inference(speech_b, input_len=None, model=self.spk_model, kwargs=kwargs, disable_pbar=True, **cfg)
                        results[_b]['spk_embedding'] = spk_res[0]['spk_embedding']
                        spk_res = self.inference(
                            speech_b, input_len=None, model=self.spk_model, kwargs=kwargs, **cfg
                        )
                        results[_b]["spk_embedding"] = spk_res[0]["spk_embedding"]
                beg_idx = end_idx
                end_idx += 1
                max_len_in_batch = sample_length
                if len(results) < 1:
                    continue
                results_sorted.extend(results)
            

            # end_asr_total = time.time()
            # time_escape_total_per_sample = end_asr_total - beg_asr_total
            # pbar_sample.update(1)
            # pbar_sample.set_description(f"rtf_avg_per_sample: {time_escape_total_per_sample / time_speech_total_per_sample:0.3f}, "
            #                      f"time_speech_total_per_sample: {time_speech_total_per_sample: 0.3f}, "
            #                      f"time_escape_total_per_sample: {time_escape_total_per_sample:0.3f}")
            

            restored_data = [0] * n
            for j in range(n):
                index = sorted_data[j][1]
                restored_data[index] = results_sorted[j]
            result = {}
            

            # results combine for texts, timestamps, speaker embeddings and others
            # TODO: rewrite for clean code
            for j in range(n):
@@ -365,12 +444,12 @@
                            t[0] += vadsegments[j][0]
                            t[1] += vadsegments[j][0]
                        result[k].extend(restored_data[j][k])
                    elif k == 'spk_embedding':
                    elif k == "spk_embedding":
                        if k not in result:
                            result[k] = restored_data[j][k]
                        else:
                            result[k] = torch.cat([result[k], restored_data[j][k]], dim=0)
                    elif 'text' in k:
                    elif "text" in k:
                        if k not in result:
                            result[k] = restored_data[j][k]
                        else:
@@ -380,66 +459,93 @@
                            result[k] = restored_data[j][k]
                        else:
                            result[k] += restored_data[j][k]
            
            return_raw_text = kwargs.get('return_raw_text', False)            

            return_raw_text = kwargs.get("return_raw_text", False)
            # step.3 compute punc model
            if self.punc_model is not None:
                self.punc_kwargs.update(cfg)
                punc_res = self.inference(result["text"], model=self.punc_model, kwargs=self.punc_kwargs, disable_pbar=True, **cfg)
                raw_text = copy.copy(result["text"])
                if return_raw_text: result['raw_text'] = raw_text
                result["text"] = punc_res[0]["text"]
                if not len(result["text"].strip()):
                    if return_raw_text:
                        result["raw_text"] = ""
                else:
                    deep_update(self.punc_kwargs, cfg)
                    punc_res = self.inference(
                        result["text"], model=self.punc_model, kwargs=self.punc_kwargs, **cfg
                    )
                    raw_text = copy.copy(result["text"])
                    if return_raw_text:
                        result["raw_text"] = raw_text
                    result["text"] = punc_res[0]["text"]
            else:
                raw_text = None
                

            # speaker embedding cluster after resorted
            if self.spk_model is not None and kwargs.get('return_spk_res', True):
            if self.spk_model is not None and kwargs.get("return_spk_res", True):
                if raw_text is None:
                    logging.error("Missing punc_model, which is required by spk_model.")
                all_segments = sorted(all_segments, key=lambda x: x[0])
                spk_embedding = result['spk_embedding']
                labels = self.cb_model(spk_embedding.cpu(), oracle_num=kwargs.get('preset_spk_num', None))
                spk_embedding = result["spk_embedding"]
                labels = self.cb_model(
                    spk_embedding.cpu(), oracle_num=kwargs.get("preset_spk_num", None)
                )
                # del result['spk_embedding']
                sv_output = postprocess(all_segments, None, labels, spk_embedding.cpu())
                if self.spk_mode == 'vad_segment':  # recover sentence_list
                if self.spk_mode == "vad_segment":  # recover sentence_list
                    sentence_list = []
                    for res, vadsegment in zip(restored_data, vadsegments):
                        if 'timestamp' not in res:
                            logging.error("Only 'iic/speech_paraformer-large-vad-punc_asr_nat-zh-cn-16k-common-vocab8404-pytorch' \
                        if "timestamp" not in res:
                            logging.error(
                                "Only 'iic/speech_paraformer-large-vad-punc_asr_nat-zh-cn-16k-common-vocab8404-pytorch' \
                                           and 'iic/speech_seaco_paraformer_large_asr_nat-zh-cn-16k-common-vocab8404-pytorch'\
                                           can predict timestamp, and speaker diarization relies on timestamps.")
                        sentence_list.append({"start": vadsegment[0],
                                              "end": vadsegment[1],
                                              "sentence": res['text'],
                                              "timestamp": res['timestamp']})
                elif self.spk_mode == 'punc_segment':
                    if 'timestamp' not in result:
                        logging.error("Only 'iic/speech_paraformer-large-vad-punc_asr_nat-zh-cn-16k-common-vocab8404-pytorch' \
                                           can predict timestamp, and speaker diarization relies on timestamps."
                            )
                        sentence_list.append(
                            {
                                "start": vadsegment[0],
                                "end": vadsegment[1],
                                "sentence": res["text"],
                                "timestamp": res["timestamp"],
                            }
                        )
                elif self.spk_mode == "punc_segment":
                    if "timestamp" not in result:
                        logging.error(
                            "Only 'iic/speech_paraformer-large-vad-punc_asr_nat-zh-cn-16k-common-vocab8404-pytorch' \
                                       and 'iic/speech_seaco_paraformer_large_asr_nat-zh-cn-16k-common-vocab8404-pytorch'\
                                       can predict timestamp, and speaker diarization relies on timestamps.")
                    sentence_list = timestamp_sentence(punc_res[0]['punc_array'],
                                                       result['timestamp'],
                                                       raw_text,
                                                       return_raw_text=return_raw_text)
                                       can predict timestamp, and speaker diarization relies on timestamps."
                        )
                    sentence_list = timestamp_sentence(
                        punc_res[0]["punc_array"],
                        result["timestamp"],
                        raw_text,
                        return_raw_text=return_raw_text,
                    )
                distribute_spk(sentence_list, sv_output)
                result['sentence_info'] = sentence_list
                result["sentence_info"] = sentence_list
            elif kwargs.get("sentence_timestamp", False):
                sentence_list = timestamp_sentence(punc_res[0]['punc_array'],
                                                   result['timestamp'],
                                                   raw_text,
                                                   return_raw_text=return_raw_text)
                result['sentence_info'] = sentence_list
            if "spk_embedding" in result: del result['spk_embedding']
                    
                if not len(result["text"].strip()):
                    sentence_list = []
                else:
                    sentence_list = timestamp_sentence(
                        punc_res[0]["punc_array"],
                        result["timestamp"],
                        raw_text,
                        return_raw_text=return_raw_text,
                    )
                result["sentence_info"] = sentence_list
            if "spk_embedding" in result:
                del result["spk_embedding"]

            result["key"] = key
            results_ret_list.append(result)
            end_asr_total = time.time()
            time_escape_total_per_sample = end_asr_total - beg_asr_total
            pbar_total.update(1)
            pbar_total.set_description(f"rtf_avg: {time_escape_total_per_sample / time_speech_total_per_sample:0.3f}, "
                                 f"time_speech: {time_speech_total_per_sample: 0.3f}, "
                                 f"time_escape: {time_escape_total_per_sample:0.3f}")

            if pbar_total:
                pbar_total.update(1)
                pbar_total.set_description(
                    f"rtf_avg: {time_escape_total_per_sample / time_speech_total_per_sample:0.3f}, "
                    f"time_speech: {time_speech_total_per_sample: 0.3f}, "
                    f"time_escape: {time_escape_total_per_sample:0.3f}"
                )

        # end_total = time.time()
        # time_escape_total_all_samples = end_total - beg_total
@@ -448,3 +554,40 @@
        #                      f"time_escape_all: {time_escape_total_all_samples:0.3f}")
        return results_ret_list

    def export(self, input=None, **cfg):
        """

        :param input:
        :param type:
        :param quantize:
        :param fallback_num:
        :param calib_num:
        :param opset_version:
        :param cfg:
        :return:
        """

        device = cfg.get("device", "cpu")
        model = self.model.to(device=device)
        kwargs = self.kwargs
        deep_update(kwargs, cfg)
        kwargs["device"] = device
        del kwargs["model"]
        model.eval()

        type = kwargs.get("type", "onnx")

        key_list, data_list = prepare_data_iterator(
            input, input_len=None, data_type=kwargs.get("data_type", None), key=None
        )

        with torch.no_grad():

            if type == "onnx":
                export_dir = export_utils.export_onnx(model=model, data_in=data_list, **kwargs)
            else:
                export_dir = export_utils.export_torchscripts(
                    model=model, data_in=data_list, **kwargs
                )

        return export_dir